Connected Classroom · Disclosure
AI Ethics Disclosure
What the tools on this site do, what they run on, and what happens to anything you type into them. Written to be checked, not to be trusted on faith.
Last updated July 2026
Who Is Behind This
Connected Classroom is the practice of one person: Timothy Cook, a practicing international classroom teacher. There is no company, no staff, and no investors. When this page says "I," that is literal.
Every commitment below is one I keep personally, so I have kept the list to commitments I can actually meet.
The Tools
The Intelligence Suite is four free tools that draft instructional options for teachers. Each one is constrained by an established pedagogical framework rather than left to improvise.
Drafts a differentiated learning menu from one objective, built on the CAST Universal Design for Learning guidelines. Options change how a student reaches the objective, never how much thinking it demands.
Turns a real student question into a scaffolded inquiry pathway. The refined question stays recognizably the student's, and every phase produces evidence the student has to interpret.
Finds the conceptual bridge between two subjects and builds the interdisciplinary lesson on it, with developmental scaffolding calibrated to the grade band.
Connects a standard to authentic real-world application. Every activity is complete inside the classroom; the community connection is enrichment, never a dependency.
Nothing here is built to reach a student directly. The tools reach children only through a teacher who has read the output and decided to use it.
The Model and the Processing
The tools run on Google's Gemini API (gemini-flash-latest), directed by system-level instructions I write and maintain, which encode the frameworks above.
Three parties touch a request, and no more. The interface is a static page served from GitHub Pages, which sees that you loaded the page but never what you type into it. Your input goes to a Google Apps Script backend I control, which holds the API credentials and the instructions so neither is ever exposed in your browser. That backend calls Gemini and returns the result. Nothing passes through any other service, and there is no analytics or advertising layer anywhere in the path.
Because the model is an alias that Google updates on its own schedule, the tools' behavior can shift without any change on my end. I re-test each tool after Gemini releases and adjust the instructions when output drifts.
Your Data
No accounts. No logins. No student data required. The tools ask for your objective or your standard. They never ask for a child's name, work, or record, and you should never enter one.
Your inputs are processed in real time and are not stored on any server I control. Anything the tool remembers between visits is saved in your own browser and clears itself after 24 hours.
I log anonymous usage counts only: which tool ran, the grade band, and the subject. Not your inputs, not the generated output, not your email, not you.
I do not sell, lease, or license anything you type. It is not used to train any model. What you make with these tools is yours, under CC BY-NC-ND 4.0, free to use in your teaching.
Full technical detail is on the Data and Governance page.
Where the Judgment Stays
- Drafts, never decisions. Every output is a starting point for your judgment, not a finished product for your class.
- The teacher is the gate. No tool here makes an automated decision about a student, and none is designed for a student to use unsupervised.
- The struggle stays. No tool answers for a child or removes the thinking a lesson exists to produce. That constraint is written into the instructions behind every tool, not added as an afterthought.
Bias, and Its Limits
Bias mitigation happens where the tool is built rather than filtered afterward. Developmental scaffolding is enforced per grade band so a K-2 output cannot arrive at 9-12 complexity. The instructions require balanced cultural, gender, and socioeconomic representation in examples, figures, and contexts. Partner and resource suggestions are constrained to institution types rather than named organizations, so the model cannot invent a real-sounding place that does not exist.
None of that makes the output unbiased. It runs on a general-purpose model trained on the open internet, and it will reproduce some of what is in there. Read every output as a draft written by someone who has never met your students.
Report a Problem
If a tool produces something inappropriate, inaccurate, or biased, write to ethics@connectedclassroom.org. Tell me which tool and what you entered, if you are comfortable sharing it.
I read every report myself and fix what I can fix. I teach full time, so I will not promise a response window I cannot keep. What I will say is that reports change the instructions behind the tools, and they have.
Standards
This framework is built against the ISTE Standards for Educators and the U.S. Department of Education's 2023 recommendations on AI and the Future of Teaching and Learning, and against the UNESCO AI competency framework.
The deeper argument for why children need different protection than adults is at the Cognitive Privacy Impact Assessment, a framework for evaluating what any AI system captures, infers, and builds dependency on before it enters a school.